Music Notes

Logo
Landing page

Inspiration

We can remember the lyrics to hundreds of songs, but never anything from the readings for class. What if we could leverage catchy music to get reading notes to stick in your head?

What it does

Music Notes will create EPs/short albums for the key ideas in a PDF, which you can listen to over and over and over again.

The user uploads a PDF file, which will be parsed for text. The key ideas of the text are grouped into topics that will each become a song on the EP, where the user can specify the genre of music they want. They can then browse their album collection for management and playback of their Music Notes.

How we built it

We're using a language model with a large context window to quickly extract the core ideas of a text and turn them into the titles and lyrics for the songs. The lyrics and genre are delivered to a music generation service that returns the audio. When the audio is available for playback, the user can listen to the track. Each collection of tracks is organized by the source PDF file so that related songs are grouped together.

Challenges we ran into

Because the text-to-song service does not have an API, we modified and used an unofficial one with FastAPI that hijacks a browser session and creates our own endpoints for generation and accessing audio files. As a result, we encountered authorization issues and other challenges related to the delayed return of the completed audio.

Accomplishments that we're proud of

There was a big lift to integrate the various APIs, especially our unofficial one. We're also pretty happy with the landing page, which had to strike the right balance between being simple and sufficiently information-dense.

What we learned

We learned about the power of friendship and enjoyed collaborating for our last WildHacks (as we're all graduating!)

What's next for Music Notes

We want to expand the kinds of inputs we can process from exclusively PDFs to other, larger documents. We also want to improve our media player so that the user can select several collections of songs to study multiple topics at once.